Innovating in Performance Management

NetScout just announced we'd been granted a new patent.  The patent, which was for a new method of anomaly detection based on statistical analysis, was the result of work by a couple of guys from Quantiva, which NetScout bought in 2005 and a mathematics/statistics consultant from Princeton.

The patented technology forms one of the key elements of the nGenius Analytics product that NetScout launched in May and is interesting because of how it improves on previous methods of doing anomaly detection for performance management in IT (as opposed to security applications, which can also use anomaly detection). Notably, using a probability distribution mapping scheme, the system is able to get a much richer picture of whether a change or trend in performance measurements is truly an anomaly, as opposed to other methods that use rolling averages or deviation-from-norm analysis which are generally more simplistic schemes.

Another interesting aspect of this patent, and part of the reason we bought Quantiva frankly, was the flexibility of patented statistical model. It doesn't make any assumptions about the data set it is analyzing (i.e. it isn't tied to Gaussian or Poisson models), which makes it inherently more useful for a variety of applications, including analyzing unpredictable network traffic. For instance, Quantiva was using the patented model for anomaly detection for a Web site performance measurement service based on response time. It could just as easily be used to detect anomalies in a stream of securities trading prices (if we ever wanted to get into that business…). Most importantly though, this flexibility will allow NetScout to apply it to potentially endless streams of network, application and other performance data, not just bandwidth usage or usage by application and response times.